www.eprg.group.cam.ac.uk

Price discrimination and limits to arbitrage in global LNG markets

EPRG Working Paper 1317

Cambridge Working Paper in Economics 1340

Robert A. Ritz

Abstract Gas prices around the world vary widely despite being connected by international trade of LNG. Some industry observers argue that major exporters (e.g., Qatar) have acted irrationally by failing to engage in price arbitrage. This is also difficult to reconcile with a perfectly competitive model in which price differences exist solely because of transport costs. We show that a model with market power can rationalize observed price differentials and trade flows. We highlight how different features of the LNG market limit the ability and/or incentive of other players to arbitrage, and discuss the potential impact of US LNG exports. Keywords International trade, limits to arbitrage, LNG pricing, market power, natural gas, price discrimination JEL Classification D40, F12, L95

Contact rar36@cam.ac.uk Publication September 2013

Price discrimination and limits to arbitrage

in global LNG markets

Robert A. Ritz∗

Faculty of Economics & EPRG

University of Cambridge

rar36@cam.ac.uk

September 2013

Abstract

Gas prices around the world vary widely despite being connected by inter-

national trade of LNG. Some industry observers argue that major exporters

(e.g., Qatar) have acted irrationally by failing to engage in price arbitrage.

This is also diffi cult to reconcile with a perfectly competitive model in which

price differences exist solely because of transport costs. We show that a model

with market power can rationalize observed price differentials and trade flows.

We highlight how different features of the LNG market limit the ability and/or

incentive of other players to arbitrage, and discuss the potential impact of US

LNG exports.

Keywords: International trade, limits to arbitrage, LNG pricing, market

power, natural gas, price discrimination

JEL classifications: D40 (Market Structure and Pricing), F12 (Trade with

Imperfect Competition), L95 (Gas Transportation)

∗Thanks are due to Alberto Behar, Chi Kong Chyong, Simon Cowan, Philipp Koenig, PeterNeary, David Newbery, Pierre Noël, Robin Smale, Yves Smeers, John Ward and Katja Yafimavafor helpful discussions and advice, to Yudi Wang and Yichi Zhang for research assistance, to Plattsfor providing some of the price data, and to participants at an Energy & Environment seminar inCambridge for their feedback. All views expressed and any errors are my own.

1

EPRG 1317

1 Introduction

Not so long ago, there was a widespread conjecture that gas prices around the

world would converge. The idea was that international trade in liquefied natural

gas (LNG) would connect previously separate geographies– notably the regional

markets of Asia, Europe and the US– and thereby link their pricing.

The volume of LNG trade has indeed grown significantly since the early 2000s,

against the backdrop of liberalization of electricity and gas industries, and a general

increase in demand for natural gas. Global investment in LNG infrastructure–

liquefaction and regasification capacity– has increased, and such facilities are now

spread across more countries. The global LNG tanker fleet has expanded signifi-

cantly, and transport costs have fallen.1

Importantly, too, contracting arrangements between buyers and sellers have be-

come more flexible. Traditionally, an LNG project involves a bilateral long-term

contract, say of a 20-year duration, between a buyer and seller, to back up the

initial investment. However, there is an ongoing shift towards trade in spot and

short-term markets; these have increased ten-fold since 2000, and now make up 25%

of total LNG sales (GIIGNL, 2012).2 This development has been aided, amongst

other things, by the adoption of Master Sales Agreements for LNG, which create

standardization and reduce transaction costs.

Yet gas prices around the world today vary widely, and these differences have

become more pronounced in the aftermath of the Fukushima accident of March

2011. The average price of natural gas in 2012 was roughly US$16/MMBtu in

Japan, around $9 in European markets like the UK, but only $3 in the US. Some

expect large regional price disparities to persist, including the International Energy

Agency in its modelling scenarios for both 2020 and 2035. In short, the gas market

today appears far from global.

For the case of the US, the reasons for price divergence are quite clear. First, the

large-scale emergence of shale gas over the last few years has put strong downward

pressure on US natural gas prices. Second, the US at present only has very limited

LNG export capability; its infrastructure still reflects the assumption of the 2000s

that the US would become a major LNG importer. As a result, the US market has

been largely isolated from the rest of the world.3

1Useful overviews of the LNG industry as of the mid 2000s are provided by Jensen (2003) andYergin and Stoppard (2003), with a focus, respectively, on economics and geopolitics.

2Brito and Hartley (2007) argue that the expectation of a shift towards short-term multilateraltrading can have self-fulfilling properties.

3Several applications to create LNG liquefaction facilities are currently pending US regula-tory/political approval.

2

EPRG 1317

The other price gaps require a different explanation, and some industry observers

have argued they imply that LNG exporters have been behaving “irrationally”.

Several major LNG exporters make short-term sales to Asia but simultaneously

supply Northwest Europe at far lower prices. This behaviour may appear irrational

in that it entails a forgone profit = |price differential| × quantity sold to the lower-priced market. For Qatari short-term sales to the UK, rather than to Japan, some

estimates suggest a forgone profit of up to $100 million per day (in late 2011), and

a cumulative figure in the billions over the period from April 2011 to April 2012.4

LNG producers are, apparently, failing to engage in price arbitrage by not exiting

the European market (at least for short-term sales).

The most immediate explanation for price divergence lies in transport costs. In

particular, a simple perfectly competitive model predicts that the price difference

across two regions served by an exporter equals the difference in the associated

transport costs. Put differently, the “netback”– that is, price minus transport cost–

for the exporter should be the same for each region.

The problem with this theory is that it cannot explain the kinds of price differ-

ences recently observed in LNG markets. Consider again the case of Qatar, which

is the largest LNG producer (with a global market share of around 30%). Figure

1 shows the differential between the Platts JKM (Japan Korea Marker) price for

Asian LNG and the UK NBP (National Balancing Point) price, plotted against the

difference in transport costs between shipping from Qatar to Japan and Qatar to

the UK, respectively.5 ,6 Prices are up to $10/MMBtu higher in Asia than in the

UK, while the corresponding transport costs are approximately identical. Perfect

competition, by contrast, predicts that these two differentials should coincide.7

The simple theory also fails to explain the data on two broader counts. First,

over large parts of the sample period (early 2010 until early 2013) it even predicts the

wrong sign: transport costs to the UK are typically slightly higher than to Japan,

while prices are much lower. Second, there is “excess volatility”: transport costs are

far too stable to be able to explain the observed volatility of (relative) gas prices.

The perfectly competitive model cannot account for other producers’behaviour.

For example, Peru is in a similar position to Qatar in that its transport-cost differ-

ential to European and Asian markets is usually very small (Platts, 2012), and yet

4These deliveries are estimated as 75% higher than contractual obligations.5Qatari sales to Japan and UK are the two largest routes of short-term LNG, with global shares

of 10% and 6% respectively (GIIGNL, 2012).6Source: Calculations based on data from Platts, Poten & Partners, and ICAP (via Bloomberg).7This theoretical prediction should, of course, not be taken too literally. Temporary deviations

are to be expected in the face of short-term demand and supply shocks. However, large pricedivergences persisting for several years are diffi cult to reconcile with this theory.

3

EPRG 1317

Figure 1: Qatar LNG sales to Japan vs UK– Differences in gas prices compared todifferences in transport costs

­2

0

2

4

6

8

10

12

2010:07 2011:01 2011:07 2012:01 2012:07 2013:01

Gas price differential (JKM ­ NBP)Transport cost differential (Japan ­ UK)

US$/MMBtu

it still makes short-term sales to both (GIIGNL, 2012). For other major producers,

such as Nigeria and Trinidad & Tobago, transport costs to Asia are indeed higher

(by around $2—3.50/MMBtu) but still often not suffi cient to explain observed sales

and prices.

In this paper, we instead suggest that regional price differentials can arise be-

cause of LNG exporters’market power. Consider a producer who can sell uncom-

mitted LNG into two export markets. In general, profit-maximization implies that

the producer equalizes marginal revenue, net of the marginal cost of production

and transport costs, across the two markets. If transport costs are identical– as is

roughly the case for Qatar’s sales to the UK and Japan– then export quantities are

such that marginal revenues for each region are equal.

Our key point is simply that equalizing marginal revenue is not necessarily the

same thing as equalizing price. Put differently, for an exporter with market power,

the “arbitrage”process stops when its marginal revenues are equalized; it is entirely

possible that this optimally leaves prices across markets far apart. This basic ar-

gument extends straightforwardly to a producer selling into more than two export

markets, and to situations with a capacity constraint on production. Moreover, the

argument does not depend on any particular assumption on the mode of (strategic)

competition in LNG; it applies for a monopolist, a Cournot-Nash player, a dominant

firm facing a competitive fringe, and so on.8

8We do not suggest that LNG producers are colluding, but rather that at least some of them

4

EPRG 1317

We provide a new general formula for relative prices between any two export

markets, in terms of a seller’s transport costs and price elasticities of demand.9

Incorporating market power can rationalize recent gas prices and trade flows by

tracing them to relative demand conditions. The Fukushima accident, for instance,

effectively switched off large parts of Japanese nuclear power, leading to an increase

in demand for imported LNG to “fill the gap”. From the viewpoint of an individual

LNG seller (with a degree of market power), under fairly general conditions, an

upward shift in market demand translates into a lower demand elasticity– and thus

a higher price. Importantly, note that local demand conditions play no role in the

competitive model, in which price differences are solely driven by transport costs.

We can also offer a perspective on the possibility that the US will become a

large LNG exporter over the coming years. What is the likely impact on gas prices?

Our analysis makes clear that US and non-US prices will not necessarily converge

as a result, even allowing for transport costs. It also suggests that any model of the

effect of US LNG exports is likely to be incomplete if it does not take market power

into account. For example, a recent model-based simulation for the US Department

of Energy incorporates general-equilibrium effects but assumes that LNG producers

do not respond strategically to US market entry.10

Our model of price discrimination with imperfect competition is static and thus

does not capture intertemporal features of the LNG market (such as storage).11 In

practice, there may be a dynamic interaction between short-term LNG prices and

long-term contracts. (Our model examines pricing incentives in short-term markets,

taking long-term commitments as given.) In particular, a large proportion of LNG

imports is still governed by long-term contracts whose terms may be renegotiated

from time to time. Some LNG producers may have been reluctant to push down

short-term prices in Asia insofar as this would make it more diffi cult to sustain

“high”prices on long-term contracts in the future. At its core, this argument has

a similar flavour to ours; while our model is based on exporters’market power in

short-term markets, this argument is essentially about exporters’bargaining power

in negotiations of long-term contracts.

Related literature. A number of empirical papers have examined price conver-

have a degree of market power (i.e., are not textbook price-takers) in some of their export markets.See Egging, Holz, von Hirschhausen and Gabriel (2009) for a recent analysis of “Gas OPEC”.

9Our approach does not rely on production cost data.10NERA (2012) assumes Qatar is “large”but does not alter its production strategy in response to

US exports, while other producers are represented as a competitive fringe; the model is augmentedwith (exogenous) mark-up adjustments in order to be able to replicate observed regional gas prices.11Chaton and Durant-Viel (2013) analyze the value of gas storage when firms have market power.

5

EPRG 1317

gence in natural gas markets.12 Siliverstovsa et al. (2005) obtain mixed results

from a cointegration analysis using data from the early 1990s until 2004; they find

evidence for market integration between European and Japanese markets, but no

integration between North America and Japan. Over the period 1999 to 2008, Neu-

mann (2009) finds increased convergence of gas spot prices between North America

and Asia. However, there does not appear to be any econometric analysis of gas-price

convergence post-2011.13 Others have focused on price convergence within regional

markets; for example, Doane and Spulber (1994) employ similar techniques to an

integrated, national market for natural gas in the US.14 By contrast, we here offer

an economic-theory perspective on price non-convergence, with a view to explaining

observed prices since March 2011. We use the standard non-cooperative approach

to pricing strategies; see Hubert and Ikkonikova (2011) for a recent application of

cooperative game theory to bargaining and investment in natural gas markets.

Also related is the literature on market power in crude oil, which estimates the

level of market competitiveness with a particular emphasis on the role of OPEC,

see, e.g., Salant (1976), Almoguera, Douglas and Herrera (2011), and Nakov and

Nuño (2013). Much of this literature concludes that the world crude oil market can

be modelled either as a set of dominant producers facing a competitive fringe, or

as something close to Cournot-Nash competition. Thereby, most oil models simply

assume a single global oil price; indeed, international price differentials in crude

oil are typically small and mainly reflect quality differences between oil varieties

from different regions. In contrast, this paper focuses on natural gas, for which the

existing literature– and the market– is much less well-developed; it obtains results

that apply for a wide range of competitive conditions, and focuses explicitly on the

limits to international price arbitrage in LNG.15

The remainder of this paper proceeds as follows. Section 2 presents a model of

LNG pricing across different export markets to formalize the market-power argument

we have outlined above. Section 3 provides a discussion of how different features of

the LNG market limit the ability and/or incentive of other players (such as LNG

buyers and third-party traders) to engage in price arbitrage, including constraints

in LNG shipping. Section 4 offers concluding remarks, including a discussion of the

12The chapters in Stern (2012) provide a detailed overview of how gas-pricing mechanisms varyacross different regions.13We conjecture that international price correlations have declined significantly (compared to

the mid/late 2000s), and that the Fukushima accident represents a structural break.14There are also several recent simulation-based models of gas-market integration in the EU, see,

e.g., Holz, von Hirschhausen and Kemfert (2008) and Lise and Hobbs (2009).15There may be an interesting analogy between the globalization of oil markets during the

1970s/1980s and ongoing developments in natural gas, but this is beyond our current scope.

6

EPRG 1317

potential effects of greater price arbitrage, in the future, on LNG prices, industry

profits, and social welfare.

2 A profit-maximizing LNG exporter

Consider an LNG producer selling output into N ≥ 2 export markets.

Let pki (xki , y

ki , X

−ki , Y −k

i ; θi) denote the inverse demand function producer k faces

in market i (i = 1, 2, ..., N), where pki is the “spot”price of LNG,16 xki is the quantity

sold by the producer in the short-term market, while yki is the quantity the producer

has pre-committed to sell in market i by way of long-term contracts (which we here

take as given). Analogously, X−ki is the vector of outputs sold by other producers in

the spot market while Y −ki captures their long-term commitments.17 Other factors

that affect demand in market i are summarized by the vector θi. For example,

this might include the prices of coal, oil and other substitute products, the state of

business cycle, other demand shocks, the weather, and so on.18

The producer’s cost function Ck(∑N

i=1(xki + yki )) depends on the sum of total

quantities sold in all N export markets, including spot market sales as well as long-

term commitments. Production may be subject to a capacity constraint such that∑Ni=1(x

ki + yki ) ≤ Q

k.19 In addition to this, the producer incurs a transport cost tki

per unit of output sold to market i. This mainly reflects the cost of shipping and

may vary across export markets depending on distance and other factors (and also

vary across different producers).

Producer k’s profit-maximization problem is to choose the amount of LNG to

export to each market, given any long-term commitments already entered into:

max{xki }Ni=1

Πk =∑N

i=1 pki x

ki − Ck(

∑Ni=1(x

ki + yki ))−

∑Ni=1 t

ki x

ki

subject to∑N

i=1(xki + yki ) ≤ Q

k.

We assume, without much loss of additional economic insight, that this problem

is well-behaved with an interior solution for each of the N export markets. The

Langrangean for constrained optimization can be written as Lk = Πk + λk(Q −16We allow prices to be producer-specific, e.g., to be able to reflect (small) quality differentials.17From a buyer’s point of view, LNG from short-term markets and long-term contracts may be

imperfect substitutes.18Some of these factors may affect individual producers in different ways, and some may influence

demand conditions in several export markets.19Adding a production constraint in terms of minimum throughput to ensure smooth operation,∑Ni=1

(xki + y

ki

)≥ Qk, would not affect the following results.

7

EPRG 1317

∑Ni=1(x

ki + yki )), where λk ≥ 0 is the shadow value of the capacity constraint (i.e.,

the value of an incremental relaxation of the capacity constraint), which is non-zero

if the producer is capacity-constrained and zero if it is not).

The optimal output choice x̂ki by producer k in market i satisfies the first-order

conditionMRki−MCk−tki−λk = 0, whereMRk

i is marginal revenue from short-term

sales, and MCk is the marginal cost of production. Using the first-order conditions

for any two export markets, say i and j, shows that these are related by

MRki − tki = MRk

j − tkj .

This is the fundamental condition for profit-maximization. The cost of an additional

unit of output is the same regardless where it ends up being sold in, both in terms

of the marginal cost of production and the shadow value of using the capacity

elsewhere. (This holds regardless of whether or not the producer is, in fact, capacity-

constrained.) To maximize profits, therefore, the producer balances at the margin

the contribution of each export market in terms of sales revenue and transport costs.

So marginal revenue net of transport costs is equalized across export markets.

Marginal revenue in market i can be written as MRki = pki

(1− 1/ηki

), where

ηki is the own-price elasticity of producer k’s demand. From here on, this elasticity

is understood to be evaluated at producer k’s optimally chosen output x̂ki , as well

as at the actual levels of short-term output of other producers and corresponding

actual long-term commitments, that is, ηki = ηki (x̂ki , y

ki , X

−ki , Y −k

i ; θi).20 Combining

this expression with the fundamental condition for profit-maximization leads to:

Proposition 1 A profit-maximizing producer k sells into N ≥ 2 export markets

with a common marginal cost (and possibly subject to a capacity constraint, Qk).

(A) In any two markets i and j, profit-maximizing prices (pki , pkj ) satisfy

(pki − pkj )pki

=ηki(

ηki − 1) [( 1

ηki− 1

ηkj

)+

(tki − tkj

)pki

],

where (tki , tkj ) are transport costs and (η

ki , η

kj ) are own-price elasticities of demand;

(B) Any observed prices (pki , pkj ) and transport costs (t

ki , t

kj ) in markets i and j can

be rationalized by some values for the price elasticities of demand (ηki , ηkj ).

Understanding profit-maximizing prices. The formula for relative prices from(A) is rather general: it does not rely on any specific functional-form assumptions

20A necessary condition for profit-maximization is that producer demand remains price-elasticin each market, ηki > 1. (Otherwise the producer could profitably reduce output.) Market-leveldemand elasticities can be significantly lower.

8

EPRG 1317

on demand and cost functions (e.g., linear, constant-elasticity, etc.), or on a partic-

ular form of competitive conduct in each export markets. Commonly-used models,

e.g., perfect competition, monopoly, Cournot-Nash oligopoly, dominant firm with a

competitive fringe, etc., are nested as special cases; the mode of competition may

also differ across export markets. An informational advantage is that it does not

feature production costs. (The result does also not assume that either consumers

or other producers are payoff-maximizers; their behaviour, rational or otherwise,

is fully captured by the producer’s own-price elasticities of demand across export

markets.)

To understand the properties of the model, consider a few special cases:

First, the simple perfectly competitive model is nested where the producer’s

demand elasticity ηki → ∞ in each export market i = 1, 2, ..., N . This corresponds

to a situation in which the producer is a price taker without any market power (so

its marginal revenue is equal to the market price in each market). Prices in any two

markets satisfy(pki − pkj

)=(tki − tkj

), and netbacks are equalized.

Second, suppose that the equilibrium values of the price elasticities of demand

in two exports market are identical, ηki = ηkj = η̂ < ∞. Then the expression forthe price differential becomes

(pki − pkj

)=(tki − tkj

)η̂/ (η̂ − 1). Relative to perfect

competition, (symmetric) market power thus exacerbates any price differential across

export markets that is due to transport costs.

Third, assume that transport costs to two markets are identical, tki = tkj . Relative

price then satisfies(pki − pkj

)/pki =

(ηkj − ηki

)/(ηki − 1

)ηkj so p

ki > pkj if and only if

ηki < ηkj . This shows that (i) prices can diverge across markets for reasons of market

power, not transport costs, and that (ii) “stronger”markets, in which a producer

faces a lower price elasticity of demand, have higher prices.

Fourth, if price elasticities and transport costs across two markets satisfy ηki ≤ ηkj

and tki ≥ tkj , then prices must satisfy pki ≥ pkj . Intuitively, market i is “far-and-

strong”, with greater market power as well as higher costs, while market j is “near-

and-weak”. If either of these relationships is strict, then pki > pkj .

A model with market power can thus explain a far greater range of observed

prices than the simple competitive model. Most importantly, transport costs are

no longer the sole driver of price differentials; relative demand conditions across

export markets now also play a key role. It relaxes the strong restriction that

sign(pki − pkj

)= sign

(tki − tkj

), and also features “excess volatility”in prices, by going

beyond the implication from the competitive model that var(pki − pkj

)= var

(tki − tkj

).

Rationalizing observed prices. Part (B) of the result is that the model withmarket power can rationalize any observed price differences between export markets.

9

EPRG 1317

The reason is as follows. Appropriate choice of the price elasticity for market i can

generate any non-negative price-cost margin, ranging from zero (when ηki →∞) toarbitrarily large (when ηki → 1), regardless of the underlying details of the producer’s

costs.21 So it is always possible to find an elasticity to rationalize the observed price

in any market, and thus also to generate correct relative prices across markets.

Consider a numerical example based on Qatari LNG sales to Japan and North-

west Europe (i = Japan, j = UK, k = Qatar). Let prices pki = 16 and pkj = 9, and

assume, for simplicity, that transport costs are identical, tki = tkj . It is not diffi cult

to check that these relative prices can be rationalized by a pair of elasticities ηki = 2

and ηkj = 9. Such prices can thus be explained by producers having relatively greater

market power in Japan.

So recent claims that LNG producers are acting irrationally by simultaneously

selling short-term cargoes to both Northwest Europe and Asian markets are not

necessarily correct. It can be entirely rational for a profit-maximizing seller to

pursue a strategy that leaves prices in Japan far higher, in response to stronger

demand. In effect, it uses sales to the UK to keep prices in Japan high.22

Why might producers have greater market power in Japan? The Fukushima

accident effectively switched off large parts of Japanese nuclear power, leading to an

increase in demand for imported LNG so as to “fill the gap”. From the viewpoint

of an individual LNG seller (with a degree of market power), under fairly general

conditions, an upward shift in market demand (captured formally by a change in

θi) translates into a lower price elasticity of demand. This, in turn, typically leads

to an increase in quantity of LNG supplied but also to an increase in its price, as is

consistent with market experience since Fukuskima.

More generally, it is frequently suggested that Asian buyers are more concerned

about “security of supply” than European buyers. This translates into a higher

willingness-to-pay for a unit of LNG and, all else equal, a lower elasticity. Further-

more, Asian buyers have fewer possibilities to substitute for LNG, notably because

of more limited access to Russian pipeline gas.

Estimating producer-specific elasticities of demand. A feature of the modelis that the pair of elasticities to rationalize the data is, in general, not unique. In the

numerical example for Qatar, the data pki = 16, pkj = 9, and tki = tkj are rationalized

for any pair of elasticities (ηki ,ηkj ) that satisfies (1 − ηki /ηkj )/(ηki − 1) = 7

9. Setting

(ηki ,ηkj ) = (2, 9) is but one solution. Loosely put, getting the relative elasticities

21A negative price-cost margin cannot be profit-maximizing in the present model.22It is also economically ineffi cient in that different consumers are paying different prices for

essentially the same good (so their marginal utilities are unequal).

10

EPRG 1317

across markets correct matters more than their absolute values.

Pinning down unique values for the elasticities requires more information and/or

additional modelling assumptions. We here suggest three possible approaches:

First, recall that the above results did not rely on production cost informa-

tion, or on any knowledge of whether the producer is, in fact, capacity-constrained.

However, if such information is available, this immediately identifies the producer-

specific demand elasticity. To see this, rewrite the first-order condition for profit-

maximization in market i, using the relationship MRki = pki

(1− 1/ηki

), to obtain

ηki = pki /[(pki − tki )− (MCk + λk)

]. In the numerical example for Qatar, knowledge

that (MCk + λk) = 6 would select the values (ηki ,ηkj ) = (2, 9) as the unique way of

rationalizing the data. Perhaps most realistically, an assessment that the producer is

not capacity-constrained, i.e., λk = 0, together with data on (pki ,tki ,MCk) identifies

ηki for each individual market.23

Second, it is possible, in principle, to estimate the elasticity for an individual mar-

kets, ηki , or the entire vector of elasticities{ηki}Ni=1, using econometric techniques.

This would require time-series data on prices and quantities in each market of in-

terest, as well as relevant control variables (probably including data on producers’

long-term contractual commitments).

Third, it may be possible to justify more specific assumptions on competitive

conduct. For instance, it is quite common in the analysis of natural gas markets to

assume Cournot-Nash competition between sellers. Then an individual producer k’s

demand elasticity ηki = ηi/ski , where ηi is the price elasticity of market demand, and

ski ∈ (0, 1) is the producer’s market share (in the appropriately defined market i).

Market share data are generally easier to obtain, and it is usually easier to estimate

a market-level elasticity than a producer-specific elasticity.

To continue our numerical example from above, suppose it is estimated that

two markets have identical price elasticities ηi = ηj = 12(say for natural gas); the

producer-specific elasticities (ηki ,ηkj ) = (2, 9) would then be generated by producer

k having market shares (ski ,skj ) = (25%, 55

9%) respectively in the two markets. Al-

ternatively, if the producer had identical market shares ski = skj = 10%, then the

corresponding market-level elasticities would be (ηi,ηj) = (15, 910

).

Unfortunately, we currently do not have suffi ciently rich data on costs, prices,

and quantities to be able to pursue either of the first two of these approaches; we

return to the impact of price discrimination under different competitive conditions

in our concluding remarks below.

23Put differently, for each seller, there are N first-order conditions but N +2 unknowns, namely,{ηki}Ni=1, MCk, and λk, but since MCk+λk is a suffi cient statistic for (MCk, λk) the system boils

down to N + 1 unknowns.

11

EPRG 1317

3 Limits to price arbitrage in LNG markets

Our model offers an explanation for why it may be optimal for prices across regions

to be different from the viewpoint of an LNG producer. Implicit in the model, as in

virtually all literature on price discrimination, is that other players do not undermine

this sales strategy. We here discuss a number of reasons, many particular to LNG

markets, that either limit the ability of other players to engage in arbitrage or create

incentives that work against pursuing arbitrage in the first place.

The textbook assumption that sustains price discrimination by a seller is that

buyers cannot engage in resale. Although there is a trend towards more flexible

LNG contracting arrangements, some “destination restrictions”appear to persist.

For example, it is said that state-controlled LNG exporters still normally restrict the

resale of their exports in a way that prevents them from being traded on commodity

exchanges. This means that some price arbitrage opportunities, if they exist, cannot

be exploited for contractual reasons.

LNG arbitrage may also be diffi cult because of limited shipping capacity.24 Al-

though there now are on the order of 400 vessels for transporting LNG, only a small

proportion of the fleet is uncommitted, in the sense of not being tied to a long-term

sales contract. Thus only few companies appear to have direct access to both un-

committed gas supplies and uncommitted LNG tankers. So an LNG buyer wishing

to engage in price arbitrage may find that the shipping market is either unable or

unwilling to provide transport at the required price.25 Note that this latter argu-

ment has some similarity to our model from above; it involves market power in the

shipping market rather than (or in addition to) the LNG market itself.26

In addition to this, there are at least two reasons to do with vertical structure

for why arbitrage, even if possible, may not be in the interest of an LNG buyer.

First, while redirecting cargo, say, from northwest Europe to Japan may promise a

higher price, it also means that the LNG buyer can no longer sell or use the gas

further downstream in the European market. So redirecting cargo may also forgo

downstream surplus, which works against the incentive to arbitrage. Second, owner-

ship arrangements along the LNG supply chain are much more complex than in any

24Our model assumes that an LNG producer takes transport costs as given when choosing its salesstrategy, and that transport is available for any desired export volume to any market. Moreover,our numerical illustrations use publicly available shipping rates as being representative.25In some cases, there may also be compability issues; not all import terminals are able to receive

deliveries from all types of LNG tankers.26There is also a potential feedback effect: Since short-term LNG typically involves longer dis-

tances than trade from long-term contracts, more price arbitrage tends to further tighten theshipping market, and may thus to some extent undermine itself. Thanks to Philipp Koenig forthis point.

12

EPRG 1317

simple model. Several LNG players hold partial (<100%) ownership stakes at var-

ious points along the supply chain– including in LNG production and liquefaction,

shipping, regasification, and downstream gas– as well as across different countries.

Put simply, a company may be an LNG seller in country A, an LNG buyer in coun-

try B, and have an infrastructure stake in country C. Such a player’s overall profit

function– and hence incentive structure– is more diffi cult to work out. However, it

seems plausible that the overall incentive sometimes works against arbitrage. In any

case, vertical issues mean that a simple comparison of netbacks may not be enough.

If neither LNG sellers nor LNG buyers have a strong incentive to engage in price

arbitrage, what about third parties such as traders? A recent industry report offers

an interesting perspective on this question: “The entry barriers to LNG trading

are surprisingly high– new entrants require more than just experienced traders and

trading systems. They must have access to cargoes, but the market’s liquidity is

typically held captive by the LNG liquefaction owners/upstream suppliers who are

understandably very reluctant to release volumes for traders to trade with. Traders

must also have access to shipping, either via owned vessels or the charter market.

Furthermore, certain ships can unload at certain terminals (e.g., many import ter-

minals cannot accommodate Q-Max vessels). This can make it even more diffi cult

to effi ciently connect volumes to buyers.” (JP Morgan Cazenove, 2012). That is,

physical arbitrage requires suffi cient capacities along the entire supply chain; almost

by definition, this is more diffi cult for third parties to secure.

It is worth highlighting a few other considerations which, in practice, make LNG

arbitrage diffi cult and financially risky– and are typically neglected in models of

price discrimination. The first is units. While the flow of gas is, in some sense,

continuous, the economics of LNG transport involves an indivisibility: the unit of

account is, in effect, a tanker. As a result, only players with suffi ciently “deep

pockets” can enter the market. The second is time. It can take two weeks, for

example, to ship LNG from Qatar to Japan. Given the volatility of gas prices, it

is possible for there to be a significant shift in relative prices over such a period of

time. So risk management becomes an important factor, both for LNG sales and

potential arbitrage activity. Although financial instruments for natural gas exist,

the derivatives market specifically for LNG is relatively underdeveloped at present.

Financial arbitrage, in general, can also be affected by the existence of agency costs

and capital constraints (Shleifer and Vishny, 1997).

Finally, the extent of price arbitrage in international gas markets may be limited

because arbitrageurs themselves have a degree of market power. This can result

from a combination of the “lumpiness”of LNG trade and barriers to entry discussed

13

EPRG 1317

above. In such cases, the optimal way to exploit a profitable trading opportunity

does not lead to prices being equalized, precisely because the arbitrageur realizes that

her actions have a non-zero effect on prices.27 For example, the optimal arbitrage

strategy α̂ ≡ arg maxα {α [pi (−α)− pj(α)]} of buying α units in low-price marketj to sell in high-price market i leaves pi (−α̂) 6= pj(α̂) whenever the arbitrageur has

market power in at least one of the two markets.

Taken together, these arguments suggests that, there are significant limits to the

scale of arbitrage activity which mean that gas price differentials– perhaps due to

market power in the LNG supply chain– have persisted.

4 Concluding remarks

Despite being connected by international trade in LNG, gas prices around the world

vary widely. It is particularly surprising that large price differentials have persisted

for several years now, notably since the Fukushima accident of March 2011. Some

industry observers have thus claimed that LNG producers are behaving irrationally

by failing to engage in international price arbitrage. Such relative prices are also

diffi cult to reconcile with a perfectly competitive model in which price differences

arise solely due to transport costs.

This paper presents the first attempt in the literature to address this puzzle.

It shows that observed prices and trade flows can be explained by LNG producers

having a degree of market power. Arbitrage by a profit-maximizing exporter takes

place by comparing marginal revenue across markets rather than only price. Differ-

ences in local demand conditions can leave prices far apart. We have argued that,

in addition, a combination of incentives, market power, and other constraints tends

to work against international arbitrage by LNG buyers and third-party traders.

So is gas a global market? This is partly a matter of definition. Yes, in the sense

that several LNG exporters sell into almost all major markets (except the US), and

thus connect their pricing– albeit imperfectly. No, in that there is currently no clear

tendency towards a single uniform gas price (even adjusted for transport costs).

Looking ahead, a number of recent developments, on balance, suggest that the

gas market may become (even) more global. Significant low-cost capacity may

emerge in form of LNG exports based on US shale gas. Yet other LNG projects,

notably in Australia, have higher-than-projected costs which may dampen future

27See also Borenstein, Bushnell, Knittel and Wolfram (2008) on pricing in Californian electricitymarkets around the time of the Enron collapse; they show that arbitrage oppurtunities existedbetween spot and forward markets but suggest these were left unexploited due to a combinationof market power and arbitrageurs’fear of regulatory penalties.

14

EPRG 1317

supply. Production is also becoming more flexible; floating liquefaction plant and

tankers with onboard regasification capabilities should make output more respon-

sive to relative prices, and recent plans (e.g., in Japan) to introduce LNG futures

contracts would facilitate hedging and arbitrage.

A natural question therefore is, how will greater price arbitrage affect the global

LNGmarket? The existing theoretical literature on third-degree price discrimination

offers some partial answers.28 It focuses on the effect of moving from “uniform

pricing”, where firms are forced (e.g., by regulation) to set identical prices in all

markets, to price discrimination, where there are no such constraints on relative

prices. Turned on its head, it therefore addresses the impact of an extreme scenario:

moving from unconstrained price discrimination by LNG exporters to a world with

perfect, costless arbitrage and a single gas price.

Much of the literature focuses on the case of a monopoly selling into two separate

markets with different demand conditions (but identical marginal cost); see Aguirre,

Cowan and Vickers (2010) for a recent analysis. Under fairly mild conditions, the

resulting uniform price lies between the high and low prices under discrimination.

Moreover, price discrimination is usually associated with lower aggregate consumer

surplus (across both markets)– although there are exceptions (Cowan, 2012).29

By revealed preference, moving to perfect arbitrage makes a monopolist worse off.

The situation is more complex for price-setting oligopoly, and the literature high-

lights the possibility that price discrimination may reduce industry profits (Corts,

1998). So it is at least conceivable that a shift to a global gas price might be positive

for LNG exporters (as a group). The impact of price discrimination on social welfare

is, in general, ambiguous, and depends, amongst other things, on the fine details

of the demand conditions across different markets.30 In the monopoly case, price

discrimination is often welfare-reducing– but it is probably more likely to increase

welfare under oligopoly. So a move to perfect arbitrage may actually cause global

welfare to fall; in any case, it is clear that important distributional effects arise.

However, the assumptions made to obtain these results limit their applicability

to LNG markets. First, virtually all of the existing literature focuses on monopoly

or price-setting duopoly, neither of which seems a natural choice for LNG mar-

28Stole (2007) provides a useful overview of this literature.29A smaller number of papers examine third-degree price discrimination by price-setting

oligopolies with differentiated products. With symmetric firms, the basic insights from themonopoly case carry over (Holmes, 1989). However, a richer range of outcomes is possible iffirms are asymmetric in that they do not rank different markets in the same way, that is, a marketis regarded as “strong”by one firm but as “weak”by another firm (Corts, 1998). It is then possiblethat price discrimination causes prices in both markets to move in the same direction.30See Bergemann, Brooks and Morris (2013) for a novel welfare analysis for monopoly.

15

EPRG 1317

kets. Second, most papers simply assume that firms supply all markets regardless of

the degree of price discrimination; this precludes the possibility, for example, that

greater price arbitrage might lead to some markets becoming so unattractive to LNG

exporters that they are no longer served.31 Third, it is typically assumed that each

producer has the same marginal cost for each market; this effectively rules out the

existence of transport costs, which almost inevitably vary across markets.32 Fourth,

particular features of LNG market such as the existence of long-term contracting

commitments and its complex supply chain and ownership structure are not mod-

elled. Finally, from a dynamic perspective, the higher profits that firms may derive

from the ability to price discriminate can increase their incentives to invest in LNG

infrastructure in the first place.

LNG markets seem a fruitful area for research, given their increasing importance

and the relative scarcity of existing literature. It would be useful to have more formal

results frommodels of price discrimination with more realistic market structures that

can be applied to LNG markets– and elsewhere. It would be particularly interesting

to combine economic theory with more extensive market data.

References

Acharya, Viral V., Lars A. Lochstoer and Tarun Ramadorai (2013). Limits to Ar-

bitrage and Hedging: Evidence from Commodity Markets. Journal of Financial

Economics, forthcoming.

Aguirre, Iñaki, Simon Cowan and John Vickers (2010). Monopoly Price Discrimina-

tion and Demand Curvature. American Economic Review 100, 1601—1615.

Almoguera, Pedro A., Christopher C. Douglas and Ana María Herrera (2011). Test-

ing for the Cartel in OPEC: Non-Cooperative Collusion or just Non-Cooperative?

Oxford Review of Economic Policy 27, 144—168.

Bergemann, Dirk, Benjamin Brooks and Stephen Morris (2013). The Limits of Price

Discrimination. Princeton Economic Theory Center Working Paper 052-2013, May

2013.31An exception is Layson (1994) who analyzes market opening in the monopoly case.32We conjecture that greater price arbitrage can have important implications for cost effi ciency,

both in terms of production costs and transport costs. A recent paper by Chen and Schwartz(2013) shows that price discrimination is more likely to raise welfare in the monopoly case whenthere are marginal-cost asymmetries across markets. Thanks to Simon Cowan for the pointer.

16

EPRG 1317

Borenstein, Severin, James Bushnell, Christopher R. Knittel and Catherine Wol-

fram (2008). Ineffi ciencies and Market Power in Financial Arbitrage: A Study of

California’s Electricity Markets. Journal of Industrial Economics 56, 347—378.

Brito, Dagobert L. and Peter R. Hartley (2007). Expectations and the Evolving

World Gas Market. The Energy Journal 28, 1—24.

Chaton, Corinne and Laure Durant-Viel (2013). Real Asset Valuation Under Im-

perfect Competition: Can We Forget About Market Fundamentals? Journal of

Economics & Management Strategy 22, 125—139.

Chen, Yongmin and Marius Schwartz (2013). Differential Pricing When Costs Differ:

A Welfare Analysis. Georgetown Economics Working Paper 13-01, April 2013.

Corts, Kenneth S. (1998). Third-Degree Price Discrimination in Oligopoly: All-Out

Competition and Strategic Commitment. RAND Journal of Economics 29, 306—323.

Cowan, Simon (2012). Third-Degree Price Discrimination and Consumer Surplus.

Journal of Industrial Economics 60, 333—345.

Doane, Michael J. and Daniel F. Spulber (1994). Open Access and the Evolution of

the U.S. Spot Market for Natural Gas. Journal of Law and Economics 37, 477—517.

Egging, Ruud, Franziska Holz, Christian von Hirschhausen and Steven Gabriel

(2009). Representing GASPEC with the World Gas Model. The Energy Journal

2009 Special Issue: World Natural Gas Markets and Trade: A Multi-Modelling

Perspective, 97—117.

GIIGNL (2012). The LNG Industry in 2011. International Group of Liquefied Nat-

ural Gas Importers, Paris.

Holmes, Thomas J. (1989). The Effects of Third-Degree Price Discrimination in

Oligopoly. American Economic Review 79, 244—250.

Holz, Franziska, Christian von Hirschhausen and Claudia Kemfert (2008). A Strate-

gic Model of European Gas Supply (GASMOD). Energy Economics 30, 766—788.

Hubert, Franz and Svetlana Ikonnikova (2011). Investment Options and Bargaining

Power in the Eurasian Supply Chain for Natural Gas. Journal of Industrial Eco-

nomics 59, 85—116.

Jensen, James T. (2003). The LNG Revolution. The Energy Journal 24, 1—45.

17

EPRG 1317

JP Morgan Cazenove (2012). Global LNG. Global Equity Research, London, 13

January 2012.

Layson, Stephen K. (1994). Market Opening under Third-Degree Price Discrimina-

tion. Journal of Industrial Economics 42, 335—340.

Lise, Wietze and Benjamin Hobbs (2009). A Dynamic Simulation of Market Power

in the Liberalized European Natural Gas Market. The Energy Journal 2009 Special

Issue: World Natural Gas Markets and Trade: A Multi-Modelling Perspective, 119—

136.

Nakov, Anton and Galo Nuño (2013). Saudi Arabia and the Oil Market. The Eco-

nomic Journal, forthcoming.

NERA (2012). Macroeconomic Impacts of LNG Exports from the United States.

Report for US Department of Energy, Washington DC, 3 December 2012.

Neumann, Anne (2009). Linking Natural Gas Markets — Is LNG Doing Its Job?

The Energy Journal 2009 Special Issue: World Natural Gas Markets and Trade: A

Multi-Modelling Perspective, 187—199.

Platts (2012). LNG Daily, Volume 9, Issue 197, 11 October 2012.

Salant, Stephen W. (1976). Exhaustible Resources and Industrial Structure: A

Nash—Cournot Approach to the World Oil Market. Journal of Political Economy

84, 1079—1094.

Shleifer, Andrei and Robert Vishny (1997). The Limits of Arbitrage. Journal of

Finance 52, 35—55.

Siliverstovsa, Boriss, Guillaume L’Hégaret, Anne Neumann and Christian von

Hirschhausen (2005). International Market Integration for Natural Gas? A Cointe-

gration Analysis of Prices in Europe, North America and Japan. Energy Economics

27, 603—615.

Stern, Jonathan P. (editor) (2012). The Pricing of Internationally Traded Gas. Ox-

ford: Oxford Institute for Energy Studies.

Stole, Lars (2007). Price Discrimination and Competition. In Mark Armstrong and

Robert Porter, Handbook of Industrial Organization, Volume 3. Amsterdam: Else-

vier, 2221—2299.

Yergin, Daniel and Michael Stoppard (2003). The Next Prize. Foreign Affairs 82,

103—114.

18

EPRG 1317